Mathematical analysis of regional geological hazard early warning model

The regional geological hazard early warning modeling was supposed to relate early warning with geological environment factors, or geological environment comprehensive factors, and meteorological rainfall factors. Performance and efficiency of the model were affected by raster unit selection, independence of the involved factors and significance of their correlation. The strategy of raster unit´s successive screening could improve the computing efficiency, making a big drop in necessary computation amount in the tests. If the involved factors were not independent, there would be more dispersion and less accuracy in weights of the comprehensive indexes of geo-environment. Therefore, spatial independence test, based on mathematical statistics χ² test and integrated with spatial corresponding pattern algorithm, was effective to test the spatial independence in the calculation examples, and could be applied in some spatial independence test problems. The process to establish the correlation of early warning index and geological environment factors or geological environment comprehensive factors, meteorological rainfall factors was from “constant coefficient linear correlation”, then to “linear correlation with varying coefficients”, and to “non-linear correlation” in the end. Linear significance check ( F test ) and parameter significance check ( t test ) were essential to constant coefficient linear correlation. Correlation check was necessary for the linear correlation with varying coefficients.